n8n connects AI agents to real business workflows, self-hosted if you want
November 19, 2026

n8n builds language models, agents and MCP servers into visual workflows. What the Berlin-based platform can do, what it costs and what its fair-code licence means.
What this is about
n8n is a workflow automation platform from the Berlin-based company n8n GmbH. In a visual interface you connect services such as email, CRM, spreadsheets or ticketing systems into processes that run automatically. For several years n8n has been building AI components directly into these processes: agents, language models from various providers, memory, vector databases and MCP servers.
The GitHub repository n8n-io/n8n had around 211,000 stars on 19 November 2026, and version 2.48.6 was released the same day. For teams that want to use AI not as a chat window but as part of their processes, n8n is one of the most relevant tools, not least because it can be self-hosted.
What n8n actually does
An n8n workflow is made of nodes. A trigger starts the process, for example a new email, a webhook or a schedule. Steps follow: fetch data, filter it, transform it, send it to other services. If you need more control, you can write your own JavaScript or Python code inside nodes.
For AI, according to its documentation, n8n offers nodes for language models from providers such as OpenAI, Anthropic and Google, plus agents with tools, memory and connections to vector databases. Technically, n8n builds on concepts from LangChain. MCP servers can be connected to an agent with one click. There are also evaluations to test whether an AI workflow works reliably.
If you choose, you install n8n yourself via Docker, Docker Compose, npm or a one-line script. Without a licence key, the free Community Edition runs.
Why it matters
Many AI projects fail not because of the model but because of the plumbing: where does the data come from, where should the result go, who checks it? This is exactly where n8n comes in. An agent can read data from a CRM, draft a reply and place it in a team channel for approval, all in one visible process.
According to the pricing page, all cloud plans include unlimited users, workflows and integrations. Billing is based on workflow executions, regardless of the number of steps. The Starter plan costs 20 euros a month billed annually for 2,500 executions, Pro costs 50 euros for 10,000 executions. Business costs 667 euros a month for 40,000 executions and is self-hosted according to the pricing page; Enterprise is negotiated individually. Important for privacy: the self-hosted version can run on your own infrastructure, and with local models the data can stay inside your own network.
In plain language
n8n is like a model railway with switches. Each wagon is a piece of information, each switch a rule: "Invoice? Left to accounting. Complaint? Right to support." The AI is a signal operator who can open, read and sort wagons. But you lay the tracks yourself, and you can always see where a wagon is.
A practical example
An online retailer receives around 300 customer emails a day. An n8n workflow fetches every new email, has a language model identify whether it concerns shipping, returns or invoices, and suggests a reply. For shipping questions, the workflow also checks the parcel status in the shop system.
The draft does not go straight to the customer but into a team channel. An employee approves it with one click or corrects it. After two weeks, the evaluation shows that around 60 per cent of drafts are approved unchanged. At 300 emails a day that adds up to roughly 9,000 executions a month, which fits into the Pro plan; the costs for the language model come on top.
Scope and limits
- Not an open-source licence in the strict sense: n8n is published under the Sustainable Use License, which n8n itself calls "fair-code". Use is permitted for internal business purposes as well as personal and non-commercial purposes. Anyone who wants to resell n8n as a service to third parties needs a different agreement. Parts with ".ee" in their name are enterprise code.
- Automation amplifies mistakes: an AI agent that answers emails or changes records can produce hundreds of errors in a short time if its assumptions are wrong. Approval steps and testing before going live are essential.
- Self-hosting means securing it yourself: your own n8n instance stores credentials for many services. It must be kept up to date, protected with strong authentication and not be exposed publicly without need.
n8n is also no substitute for careful process design. If you do not know how a process should work, a workflow only gives you a faster mess.
SEO & GEO keywords
n8n, workflow automation, AI agents, AI workflow automation, self-hosted, Sustainable Use License, fair-code, LangChain, MCP, no-code, low-code, n8n GmbH Berlin
💡 In plain English
n8n is a tool that runs processes between different programs automatically, for example from an email inbox into a customer system. AI can read, sort and draft replies inside these processes. You can rent n8n from n8n or run it for free on your own servers.
Key Takeaways
- →n8n is a workflow platform from Berlin-based n8n GmbH with a visual interface and code nodes.
- →AI nodes support models from OpenAI, Anthropic, Google and others, plus agents, memory, vector databases and MCP servers.
- →Version 2.48.6 was released on 19 November 2026.
- →Cloud plans start at 20 euros a month billed annually and charge per workflow execution.
- →The Community Edition can be self-hosted for free.
- →The Sustainable Use License is not a classic open-source licence and restricts reselling n8n as a service.
FAQ
Is n8n open source?
Not in the strict sense. The source code is public but published under the Sustainable Use License, which n8n calls fair-code and which restricts commercial resale.
How much does n8n cost?
The self-hosted Community Edition is free. According to the pricing page, cloud plans start at 20 euros a month billed annually for 2,500 executions.
Can n8n work with local AI models?
Yes, self-hosted instances can connect local models so that data can stay in your own network. Setting this up requires technical knowledge, though.
Do I need programming skills?
Not for simple workflows. For complex data processing and running it securely yourself, technical knowledge helps.